Shapley variable importance cloud for machine-learning-assisted cognitive assessment in people with Parkinson disease

  • Mild cognitive impairment (MCI) is common in people with Parkinson disease (PD); however, there is a lack of evidence assessing variable importance and new blood biomarkers associated with PD-MCI.
  • This study used the Shapley variable importance cloud (ShapleyVIC) that identified 22 significantly important variables associated with PD-MCI. In addition, 8 variables (representing education, hypertension, motor score, triglycerides, apolipoprotein A1, and SNCA rs6826785 noncarrier status) that were included in the final model, were significantly associated with an increased risk of PD-MCI (P<0.05).
  • The authors concluded that the ShapleyVIC proved to be a novel and robust machine learning tool that can be applied to future clinical trials for the prioritization of important variables associated with PD-MCI.